Machine learning approaches and databases for prediction of drug-target interaction: a survey paper

The task of predicting the interactions between drugs and targets plays a key role in the process of drug discovery. There is a need to develop novel and efficient prediction approaches in order to avoid costly and laborious yet not-always-deterministic experiments to determine drug–target interactions (DTIs) by experiments alone. These approaches should be capable of identifying the potential DTIs in a timely manner. In this article, we describe the data required for the task of DTI prediction followed by a comprehensive catalog consisting of machine learning methods and databases, which have been proposed and utilized to predict DTIs. The advantages and disadvantages of each set of methods are also briefly discussed. Lastly, the challenges one may face in prediction of DTI using machine learning approaches are highlighted and we conclude by shedding some lights on important future research directions.

Big Data Analytics in Healthcare

The rapidly expanding field of big data analytics has started to play a pivotal role in the evolution of healthcare practices and research. It has provided tools to accumulate, manage, analyze, and assimilate large volumes of disparate, structured, and unstructured data produced by current healthcare systems. Big data analytics has been recently applied towards aiding the process of care delivery and disease exploration. However, the adoption rate and research development in this space is still hindered by some fundamental problems inherent within the big data paradigm. In this paper, we discuss some of these major challenges with a focus on three upcoming and promising areas of medical research: image, signal, and genomics based analytics. Recent research which targets utilization of large volumes of medical data while combining multimodal data from disparate sources is discussed. Potential areas of research within this field which have the ability to provide meaningful impact on healthcare delivery are also examined.

Deep learning in pharmacogenomics: from gene regulation to patient stratification

This Perspective provides examples of current and future applications of deep learning in pharmacogenomics, including: identification of novel regulatory variants located in noncoding domains of the genome and their function as applied to pharmacoepigenomics; patient stratification from medical records; and the mechanistic prediction of drug response, targets and their interactions. Deep learning encapsulates a family of machine learning algorithms that has transformed many important subfields of artificial intelligence over the last decade, and has demonstrated breakthrough performance improvements on a wide range of tasks in biomedicine. We anticipate that in the future, deep learning will be widely used to predict personalized drug response and optimize medication selection and dosing, using knowledge extracted from large and complex molecular, epidemiological, clinical and demographic datasets.

Fully automated endoscopic disease activity assessment in ulcerative colitis

BACKGROUND AND AIMS

Endoscopy is essential for disease assessment in ulcerative colitis (UC), but subjectivity threatens accuracy and precision. We aimed to pilot a fully automated video analysis system for grading endoscopic disease in UC.

METHODS

A developmental set of high-resolution UC endoscopic videos were assigned Mayo endoscopic scores (MESs) provided by 2 experienced reviewers. Video still-image stacks were annotated for image quality (informativeness) and MES. Models to predict still-image informativeness and disease severity were trained using convolutional neural networks. A template-matching grid search was used to estimate whole-video MESs provided by human reviewers using predicted still-image MES proportions. The automated whole-video MES workflow was tested using unaltered endoscopic videos from a multicenter UC clinical trial.

RESULTS

The developmental high-resolution and testing multicenter clinical trial sets contained 51 and 264 videos, respectively. The still-image informative classifier had excellent performance with a sensitivity of 0.902 and specificity of 0.870. In high-resolution videos, fully automated methods correctly predicted MESs in 78% (41 of 50, κ = 0.84) of videos. In external clinical trial videos, reviewers agreed on MESs in 82.8% (140 of 169) of videos (κ = 0.78). Automated and central reviewer scoring agreement occurred in 57.1% of videos (κ = 0.59), but improved to 69.5% (107 of 169) when accounting for reviewer disagreement. Automated MES grading of clinical trial videos (often low resolution) correctly distinguished remission (MES 0,1) versus active disease (MES 2,3) in 83.7% (221 of 264) of videos.

CONCLUSIONS

These early results support the potential for artificial intelligence to provide endoscopic disease grading in UC that approximates the scoring of experienced reviewers.

Osteoarthritis of the Temporomandibular Joint can be diagnosed earlier using biomarkers and machine learning

After chronic low back pain, Temporomandibular Joint (TMJ) disorders are the second most common musculoskeletal condition affecting 5 to 12% of the population, with an annual health cost estimated at $4 billion. Chronic disability in TMJ osteoarthritis (OA) increases with aging, and the main goal is to diagnosis before morphological degeneration occurs. Here, we address this challenge using advanced data science to capture, process and analyze 52 clinical, biological and high-resolution CBCT (radiomics) markers from TMJ OA patients and controls. We tested the diagnostic performance of four machine learning models: Logistic Regression, Random Forest, LightGBM, XGBoost. Headaches, Range of mouth opening without pain, Energy, Haralick Correlation, Entropy and interactions of TGF-β1 in Saliva and Headaches, VE-cadherin in Serum and Angiogenin in Saliva, VE-cadherin in Saliva and Headaches, PA1 in Saliva and Headaches, PA1 in Saliva and Range of mouth opening without pain; Gender and Muscle Soreness; Short Run Low Grey Level Emphasis and Headaches, Inverse Difference Moment and Trabecular Separation accurately diagnose early stages of this clinical condition. Our results show the XGBoost + LightGBM model with these features and interactions achieves the accuracy of 0.823, AUC 0.870, and F1-score 0.823 to diagnose the TMJ OA status. Thus, we expect to boost future studies into osteoarthritis patient-specific therapeutic interventions, and thereby improve the health of articular joints.

Polyp Segmentation in Colonoscopy Images Using Fully Convolutional Network

Colorectal cancer is one of the highest causes of cancer-related death, especially in men. Polyps are one of the main causes of colorectal cancer, and early diagnosis of polyps by colonoscopy could result in successful treatment. Diagnosis of polyps in colonoscopy videos is a challenging task due to variations in the size and shape of polyps. In this paper, we proposed a polyp segmentation method based on the convolutional neural network. Two strategies enhance the performance of the method. First, we perform a novel image patch selection method in the training phase of the network. Second, in the test phase, we perform effective post-processing on the probability map that is produced by the network. Evaluation of the proposed method using the CVC-ColonDB database shows that our proposed method achieves more accurate results in comparison with previous colonoscopy video-segmentation methods.

Melanoma detection by analysis of clinical images using convolutional neural network

Melanoma, most threatening type of skin cancer, is on the rise. In this paper an implementation of a deep-learning system on a computer server, equipped with graphic processing unit (GPU), is proposed for detection of melanoma lesions. Clinical (non-dermoscopic) images are used in the proposed system, which could assist a dermatologist in early diagnosis of this type of skin cancer. In the proposed system, input clinical images, which could contain illumination and noise effects, are preprocessed in order to reduce such artifacts. Afterward, the enhanced images are fed to a pre-trained convolutional neural network (CNN) which is a member of deep learning models. The CNN classifier, which is trained by large number of training samples, distinguishes between melanoma and benign cases. Experimental results show that the proposed method is superior in terms of diagnostic accuracy in comparison with the state-of-the-art methods.

Breast cancer detection in gadolinium-enhanced MR images by static region descriptors and neural networks

PURPOSE

To automate the diagnosis of malignancy by classifying breast tissues as negative or positive for malignancy in gadolinium-enhanced dynamic magnetic resonance (MR) images, using static region descriptors and a neural network classifier.

MATERIALS AND METHODS

We propose a novel approach whereby the classifier evaluates a number of parameters that identify important tumor characteristics, as obtained by digital image processing techniques. These parameters include static signal intensity (SI) after contrast enhancement, mass margin descriptors, evaluation of mass shape by calculation of eccentricity, mass size, and mass granularity by texture analysis. Datasets for 14 patients were obtained by use of the 1.5T PMRTOW Clinical Imager.

RESULTS

Statistical performance evaluation of the neural networks indicated 90%-100% sensitivity, 91%-100% specificity, and 91%-100% accuracy.

CONCLUSION

Although this work is preliminary, it may reduce overall health-care time and costs, and enable higher accuracy in automated breast cancer detection systems.

Fetal Ultrasound Image Segmentation for Measuring Biometric Parameters Using Multi-Task Deep Learning

Ultrasound imaging is a standard examination during pregnancy that can be used for measuring specific biometric parameters towards prenatal diagnosis and estimating gestational age. Fetal head circumference (HC) is one of the significant factors to determine the fetus growth and health. In this paper, a multi-task deep convolutional neural network is proposed for automatic segmentation and estimation of HC ellipse by minimizing a compound cost function composed of segmentation dice score and MSE of ellipse parameters. Experimental results on fetus ultrasound dataset in different trimesters of pregnancy show that the segmentation results and the extracted HC match well with the radiologist annotations. The obtained dice scores of the fetal head segmentation and the accuracy of HC evaluations are comparable to the state-of-the-art.

Biomedical informatics for computer-aided decision support systems: a survey

The volumes of current patient data as well as their complexity make clinical decision making more challenging than ever for physicians and other care givers. This situation calls for the use of biomedical informatics methods to process data and form recommendations and/or predictions to assist such decision makers. The design, implementation, and use of biomedical informatics systems in the form of computer-aided decision support have become essential and widely used over the last two decades. This paper provides a brief review of such systems, their application protocols and methodologies, and the future challenges and directions they suggest.

Automated ventricular systems segmentation in brain CT images by combining low-level segmentation and high-level template matching

Background

Accurate analysis of CT brain scans is vital for diagnosis and treatment of Traumatic Brain Injuries (TBI). Automatic processing of these CT brain scans could speed up the decision making process, lower the cost of healthcare, and reduce the chance of human error. In this paper, we focus on automatic processing of CT brain images to segment and identify the ventricular systems. The segmentation of ventricles provides quantitative measures on the changes of ventricles in the brain that form vital diagnosis information.

Methods

First all CT slices are aligned by detecting the ideal midlines in all images. The initial estimation of the ideal midline of the brain is found based on skull symmetry and then the initial estimate is further refined using detected anatomical features. Then a two-step method is used for ventricle segmentation. First a low-level segmentation on each pixel is applied on the CT images. For this step, both Iterated Conditional Mode (ICM) and Maximum A Posteriori Spatial Probability (MASP) are evaluated and compared. The second step applies template matching algorithm to identify objects in the initial low-level segmentation as ventricles. Experiments for ventricle segmentation are conducted using a relatively large CT dataset containing mild and severe TBI cases.

Results

Experiments show that the acceptable rate of the ideal midline detection is over 95%. Two measurements are defined to evaluate ventricle recognition results. The first measure is a sensitivity-like measure and the second is a false positive-like measure. For the first measurement, the rate is 100% indicating that all ventricles are identified in all slices. The false positives-like measurement is 8.59%. We also point out the similarities and differences between ICM and MASP algorithms through both mathematically relationships and segmentation results on CT images.

Conclusion

The experiments show the reliability of the proposed algorithms. The novelty of the proposed method lies in its incorporation of anatomical features for ideal midline detection and the two-step ventricle segmentation method. Our method offers the following improvements over existing approaches: accurate detection of the ideal midline and accurate recognition of ventricles using both anatomical features and spatial templates derived from Magnetic Resonance Images.

A Review of Automated Methods for Detection of Myocardial Ischemia and Infarction Using Electrocardiogram and Electronic Health Records

There is a growing body of research focusing on automatic detection of ischemia and myocardial infarction (MI) using computer algorithms. In clinical settings, ischemia and MI are diagnosed using electrocardiogram (ECG) recordings as well as medical context including patient symptoms, medical history, and risk factors-information that is often stored in the electronic health records. The ECG signal is inspected to identify changes in the morphology such as ST-segment deviation and T-wave changes. Some of the proposed methods compute similar features automatically while others use nonconventional features such as wavelet coefficients. This review provides an overview of the methods that have been proposed in this area, focusing on their historical evolution, the publicly available datasets that they have used to evaluate their performance, and the details of their algorithms for ECG and EHR analysis. The validation strategies that have been used to evaluate the performance of the proposed methods are also presented. Finally, the paper provides recommendations for future research to address the shortcomings of the currently existing methods and practical considerations to make the proposed technical solutions applicable in clinical practice.

Accounting for Label Uncertainty in Machine Learning for Detection of Acute Respiratory Distress Syndrome

When training a machine learning algorithm for a supervised-learning task in some clinical applications, uncertainty in the correct labels of some patients may adversely affect the performance of the algorithm. For example, even clinical experts may have less confidence when assigning a medical diagnosis to some patients because of ambiguity in the patient's case or imperfect reliability of the diagnostic criteria. As a result, some cases used in algorithm training may be mislabeled, adversely affecting the algorithm's performance. However, experts may also be able to quantify their diagnostic uncertainty in these cases. We present a robust method implemented with support vector machines (SVM) to account for such clinical diagnostic uncertainty when training an algorithm to detect patients who develop the acute respiratory distress syndrome (ARDS). ARDS is a syndrome of the critically ill that is diagnosed using clinical criteria known to be imperfect. We represent uncertainty in the diagnosis of ARDS as a graded weight of confidence associated with each training label. We also performed a novel time-series sampling method to address the problem of intercorrelation among the longitudinal clinical data from each patient used in model training to limit overfitting. Preliminary results show that we can achieve meaningful improvement in the performance of algorithm to detect patients with ARDS on a hold-out sample, when we compare our method that accounts for the uncertainty of training labels with a conventional SVM algorithm.

Non-linear dynamical signal characterization for prediction of defibrillation success through machine learning

Background

Ventricular Fibrillation (VF) is a common presenting dysrhythmia in the setting of cardiac arrest whose main treatment is defibrillation through direct current countershock to achieve return of spontaneous circulation. However, often defibrillation is unsuccessful and may even lead to the transition of VF to more nefarious rhythms such as asystole or pulseless electrical activity. Multiple methods have been proposed for predicting defibrillation success based on examination of the VF waveform. To date, however, no analytical technique has been widely accepted. We developed a unique approach of computational VF waveform analysis, with and without addition of the signal of end-tidal carbon dioxide (PetCO2), using advanced machine learning algorithms. We compare these results with those obtained using the Amplitude Spectral Area (AMSA) technique.

Methods

A total of 90 pre-countershock ECG signals were analyzed form an accessible preshosptial cardiac arrest database. A unified predictive model, based on signal processing and machine learning, was developed with time-series and dual-tree complex wavelet transform features. Upon selection of correlated variables, a parametrically optimized support vector machine (SVM) model was trained for predicting outcomes on the test sets. Training and testing was performed with nested 10-fold cross validation and 6–10 features for each test fold.

Results

The integrative model performs real-time, short-term (7.8 second) analysis of the Electrocardiogram (ECG). For a total of 90 signals, 34 successful and 56 unsuccessful defibrillations were classified with an average Accuracy and Receiver Operator Characteristic (ROC) Area Under the Curve (AUC) of 82.2% and 85%, respectively. Incorporation of the end-tidal carbon dioxide signal boosted Accuracy and ROC AUC to 83.3% and 93.8%, respectively, for a smaller dataset containing 48 signals. VF analysis using AMSA resulted in accuracy and ROC AUC of 64.6% and 60.9%, respectively.

Conclusion

We report the development and first-use of a nontraditional non-linear method of analyzing the VF ECG signal, yielding high predictive accuracies of defibrillation success. Furthermore, incorporation of features from the PetCO2 signal noticeably increased model robustness. These predictive capabilities should further improve with the availability of a larger database.

A comparative analysis of multi-level computer-assisted decision making systems for traumatic injuries

BACKGROUND

This paper focuses on the creation of a predictive computer-assisted decision making system for traumatic injury using machine learning algorithms. Trauma experts must make several difficult decisions based on a large number of patient attributes, usually in a short period of time. The aim is to compare the existing machine learning methods available for medical informatics, and develop reliable, rule-based computer-assisted decision-making systems that provide recommendations for the course of treatment for new patients, based on previously seen cases in trauma databases. Datasets of traumatic brain injury (TBI) patients are used to train and test the decision making algorithm. The work is also applicable to patients with traumatic pelvic injuries.

METHODS

Decision-making rules are created by processing patterns discovered in the datasets, using machine learning techniques. More specifically, CART and C4.5 are used, as they provide grammatical expressions of knowledge extracted by applying logical operations to the available features. The resulting rule sets are tested against other machine learning methods, including AdaBoost and SVM. The rule creation algorithm is applied to multiple datasets, both with and without prior filtering to discover significant variables. This filtering is performed via logistic regression prior to the rule discovery process.

RESULTS

For survival prediction using all variables, CART outperformed the other machine learning methods. When using only significant variables, neural networks performed best. A reliable rule-base was generated using combined C4.5/CART. The average predictive rule performance was 82% when using all variables, and approximately 84% when using significant variables only. The average performance of the combined C4.5 and CART system using significant variables was 89.7% in predicting the exact outcome (home or rehabilitation), and 93.1% in predicting the ICU length of stay for airlifted TBI patients.

CONCLUSION

This study creates an efficient computer-aided rule-based system that can be employed in decision making in TBI cases. The rule-bases apply methods that combine CART and C4.5 with logistic regression to improve rule performance and quality. For final outcome prediction for TBI cases, the resulting rule-bases outperform systems that utilize all available variables.

Vessel extraction in X-ray angiograms using deep learning

Coronary artery disease (CAD) is the most common type of heart disease which is the leading cause of death all over the world. X-ray angiography is currently the gold standard imaging technique for CAD diagnosis. These images usually suffer from low quality and presence of noise. Therefore, vessel enhancement and vessel segmentation play important roles in CAD diagnosis. In this paper a deep learning approach using convolutional neural networks (CNN) is proposed for detecting vessel regions in angiography images. Initially, an input angiogram is preprocessed to enhance its contrast. Afterward, the image is evaluated using patches of pixels and the network determines the vessel and background regions. A set of 1,040,000 patches is used in order to train the deep CNN. Experimental results on angiography images of a dataset show that our proposed method has a superior performance in extraction of vessel regions.

A hierarchical expert-guided machine learning framework for clinical decision support systems: an application to traumatic brain injury prognostication

Prognosis of the long-term functional outcome of traumatic brain injury is essential for personalized management of that injury. Nonetheless, accurate prediction remains unavailable. Although machine learning has shown promise in many fields, including medical diagnosis and prognosis, such models are rarely deployed in real-world settings due to a lack of transparency and trustworthiness. To address these drawbacks, we propose a machine learning-based framework that is explainable and aligns with clinical domain knowledge. To build such a framework, additional layers of statistical inference and human expert validation are added to the model, which ensures the predicted risk score’s trustworthiness. Using 831 patients with moderate or severe traumatic brain injury to build a model using the proposed framework, an area under the receiver operating characteristic curve (AUC) and accuracy of 0.8085 and 0.7488 were achieved, respectively, in determining which patients will experience poor functional outcomes. The performance of the machine learning classifier is not adversely affected by the imposition of statistical and domain knowledge “checks and balances”. Finally, through a case study, we demonstrate how the decision made by a model might be biased if it is not audited carefully.

Signatures of tumor-immune interactions as biomarkers for breast cancer prognosis

Breast cancer mortality is usually due to distant recurrence of cancer at an advanced stage of the disease rather than from primary cancer. Therefore, prediction of breast cancer recurrence at the time of diagnosis could lead to advances in personalized treatment of cancer patients in order to prevent risk of recurrence. Two prognostic biomarkers that are currently being used in clinical practice are a 70-gene MammaPrint signature and a 21-gene Oncotype DX panel. These assays generate relative risks of recurrence, but they do not provide a 'yes' or 'no' answer about recurrence in a given patient. These tests include genes that are involved in the cell cycle, invasion, metastasis and angiogenesis related to breast cancer. Emerging evidence suggests that a signature of genes involved in tumor-immune interactions may provide a more accurate prognostic tool. This paper reviews recent advances in the discovery of prognostic biomarkers for breast cancer patients.

An automated method for analysis of microcirculation videos for accurate assessment of tissue perfusion

Background

Imaging of the human microcirculation in real-time has the potential to detect injuries and illnesses that disturb the microcirculation at earlier stages and may improve the efficacy of resuscitation. Despite advanced imaging techniques to monitor the microcirculation, there are currently no tools for the near real-time analysis of the videos produced by these imaging systems. An automated system tool that can extract microvasculature information and monitor changes in tissue perfusion quantitatively might be invaluable as a diagnostic and therapeutic endpoint for resuscitation.

Methods

The experimental algorithm automatically extracts microvascular network and quantitatively measures changes in the microcirculation. There are two main parts in the algorithm: video processing and vessel segmentation. Microcirculatory videos are first stabilized in a video processing step to remove motion artifacts. In the vessel segmentation process, the microvascular network is extracted using multiple level thresholding and pixel verification techniques. Threshold levels are selected using histogram information of a set of training video recordings. Pixel-by-pixel differences are calculated throughout the frames to identify active blood vessels and capillaries with flow.

Results

Sublingual microcirculatory videos are recorded from anesthetized swine at baseline and during hemorrhage using a hand-held Side-stream Dark Field (SDF) imaging device to track changes in the microvasculature during hemorrhage. Automatically segmented vessels in the recordings are analyzed visually and the functional capillary density (FCD) values calculated by the algorithm are compared for both health baseline and hemorrhagic conditions. These results were compared to independently made FCD measurements using a well-known semi-automated method. Results of the fully automated algorithm demonstrated a significant decrease of FCD values. Similar, but more variable FCD values were calculated using a commercially available software program requiring manual editing.

Conclusions

An entirely automated system for analyzing microcirculation videos to reduce human interaction and computation time is developed. The algorithm successfully stabilizes video recordings, segments blood vessels, identifies vessels without flow and calculates FCD in a fully automated process. The automated process provides an equal or better separation between healthy and hemorrhagic FCD values compared to currently available semi-automatic techniques. The proposed method shows promise for the quantitative measurement of changes occurring in microcirculation during injury.

Motion Artifact Suppression in Impedance Pneumography Signal for Portable Monitoring of Respiration: An Adaptive Approach

The focus of this paper is motion artifact (MA) reduction from the impedance pneumography (IP) signal, which is widely used to monitor respiration. The amplitude of the MA that contaminates the IP signal is often much larger than the amplitude of the respiratory component of the signal. Moreover, the morphology and frequency composition of the artifacts may be very similar to that of the respiration, making it difficult to remove these artifacts. The proposed filter uses a regularization term to ensure that the pattern of the filtered signal is similar to that of respiration. It also ensures that the amplitude of the filter output is within the expected range of the IP signal by imposing an ε-tube on the filtered signal. The adaptive ε-tube filter is 100 times faster than the previously proposed nonadaptive version and achieves higher accuracies. Moreover, the experimental results, using several different performance measures, suggest that the proposed method outperforms popular MA reduction methods such as normalized least mean squares (NLMS) and recursive least squares (RLS) as well as independent component analysis (ICA). When used to extract the respiratory rate, the adaptive ε-tube achieves a mean error of 1.27 breaths per minute (BPM) compared to 4.72 and 4.63 BPM for the NLMS and RLS filters, respectively. When compared to the ICA algorithm, the proposed filter has an error of 1.06 BPM compared to 3.47 BPM for ICA. The statistical analyses indicate that all of the reported performance improvements are significant.

Automated Segmentation and Severity Analysis of Subdural Hematoma for Patients with Traumatic Brain Injuries

Detection and severity assessment of subdural hematoma is a major step in the evaluation of traumatic brain injuries. This is a retrospective study of 110 computed tomography (CT) scans from patients admitted to the Michigan Medicine Neurological Intensive Care Unit or Emergency Department. A machine learning pipeline was developed to segment and assess the severity of subdural hematoma. First, the probability of each point belonging to the hematoma region was determined using a combination of hand-crafted and deep features. This probability provided the initial state of the segmentation. Next, a 3D post-processing model was applied to evolve the initial state and delineate the hematoma. The recall, precision, and Dice similarity coefficient of the proposed segmentation method were 78.61%, 76.12%, and 75.35%, respectively, for the entire population. The Dice similarity coefficient was 79.97% for clinically significant hematomas, which compared favorably to an inter-rater Dice similarity coefficient. In volume-based severity analysis, the proposed model yielded an F1, recall, and specificity of 98.22%, 98.81%, and 92.31%, respectively, in detecting moderate and severe subdural hematomas based on hematoma volume. These results show that the combination of classical image processing and deep learning can outperform deep learning only methods to achieve greater average performance and robustness. Such a system can aid critical care physicians in reducing time to intervention and thereby improve long-term patient outcomes.

Utilization of smartphone and tablet camera photographs to predict healing of diabetes-related foot ulcers

The objective of this study was to build a machine learning model that can predict healing of diabetes-related foot ulcers, using both clinical attributes extracted from electronic health records (EHR) and image features extracted from photographs. The clinical information and photographs were collected at an academic podiatry wound clinic over a three-year period. Both hand-crafted color and texture features and deep learning-based features from the global average pooling layer of ResNet-50 were extracted from the wound photographs. Random Forest (RF) and Support Vector Machine (SVM) models were then trained for prediction. For prediction of eventual wound healing, the models built with hand-crafted imaging features alone outperformed models built with clinical or deep-learning features alone. Models trained with all features performed comparatively against models trained with hand-crafted imaging features. Utilization of smartphone and tablet photographs taken outside of research settings hold promise for predicting prognosis of diabetes-related foot ulcers.